“Last-Mile” preparation for a potential disaster – Interdisciplinary

Nat. Hazards Earth Syst. Sci., 9, 1509–1528, 2009www.nat-hazards-earth-syst-sci.net/9/1509/2009/© Author(s) 2009. This work is distributed underthe Creative Commons Attribution 3.0 License.

Natural Hazardsand Earth

System Sciences

“Last-Mile” preparation for a potential disaster – Interdisciplinaryapproach towards tsunami early warning and an evacuationinformation system for the coastal city of Padang, Indonesia

H. Taubenbock1,2, N. Goseberg3, N. Setiadi4, G. Lammel5, F. Moder6, M. Oczipka7, H. Kl upfel8, R. Wahl9,T. Schlurmann3, G. Strunz1, J. Birkmann4, K. Nagel5, F. Siegert6, F. Lehmann7, S. Dech1,2, A. Gress9, and R. Klein9

1German Remote Sensing Data Center (DFD), German Aerospace Center (DLR), 82234 Oberpfaffenhofen, Germany2Julius-Maximilians University Wurzburg, Geographical Institute, Earth Observation, 97074 Wurzburg, Germany3Leibniz University Hannover, Franzius-Institut fur Wasserbau und Kusteningenieurwesen, 30167 Hannover, Germany4United Nations University, Institute for Environment and Human Security (UNU-EHS), 53113 Bonn, Germany5Technical University Berlin, Verkehrssystemplanung und Verkehrstelematik, Institut fur Land- und Seeverkehr,Fakultat V – Verkehr- und Maschinensysteme, 10587 Berlin, Germany6Remote Sensing Solutions GmbH (RSS), 81667 Munchen, Germany7Institut fur Robotik und Mechatronik, German Aerospace Center (DLR), Berlin-Adlershof, Germany8TraffGo HT GmbH, Falkstr. 73–77, 47058 Duisburg, Germany9Insitute of Computer Science II, University of Bonn, 53117 Bonn, Germany

Received: 4 May 2009 – Revised: 24 July 2009 – Accepted: 6 August 2009 – Published: 28 August 2009

Abstract. Extreme natural events, like e.g. tsunamis orearthquakes, regularly lead to catastrophes with dramaticconsequences. In recent years natural disasters caused hun-dreds of thousands of deaths, destruction of infrastructure,disruption of economic activity and loss of billions of dol-lars worth of property and thus revealed considerable deficitshindering their effective management: Needs for stakehold-ers, decision-makers as well as for persons concerned includesystematic risk identification and evaluation, a way to as-sess countermeasures, awareness raising and decision sup-port systems to be employed before, during and after crisissituations. The overall goal of this study focuses on interdis-ciplinary integration of various scientific disciplines to con-tribute to a tsunami early warning information system. Incomparison to most studies our focus is on high-end geomet-ric and thematic analysis to meet the requirements of small-scale, heterogeneous and complex coastal urban systems.Data, methods and results from engineering, remote sensingand social sciences are interlinked and provide comprehen-sive information for disaster risk assessment, managementand reduction. In detail, we combine inundation modeling,urban morphology analysis, population assessment, socio-

Correspondence to:H. Taubenbock([email protected])

economic analysis of the population and evacuation model-ing. The interdisciplinary results eventually lead to recom-mendations for mitigation strategies in the fields of spatialplanning or coping capacity.

1 Introduction

Tsunamis are an ever-present threat to lives, infrastructure,and property along the coasts of the world’s oceans. In-donesia has recently been struck by a series of major earth-quakes and arising tsunamis that have largely affected cities:the 2004 tsunami and earthquakes (devastating the citiesof Banda Aceh and Meulaboh) and the 2006 central Javaearthquake (impacting the Yogyakarta region) with an en-suing tsunami (Leitmann, 2007). The dimension of the2004 tsunami disaster claiming nearly 275 000 lives and de-stroying billions of dollars’ worth of property (Barber, 2005)reminded the world, to be more proactive in developing waysto reduce their impact.

Today, several end-to-end tsunami early warning sys-tems are operated. Examples are the US National TsunamiHazard Mitigation System (NTHMP) in the eastern Pacific(Bernhard, 2005; Titov et al., 2005), the Japanese UrgentEarthquake Detection and Alarm System (UrEDAS) in thewestern Pacific (Saita and Nakamura, 2003) or the German

Published by Copernicus Publications on behalf of the European Geosciences Union.

http://creativecommons.org/licenses/by/3.0/

1510 H. Taubenbock et al.: “Last-Mile” preparation to a potential disaster

Indonesian Early Warning System (GITEWS) in the IndianOcean (Lauterjung, 2005). The manifold aspects for end-to-end tsunami early warning systems include tsunami hazardand vulnerability assessment, perception studies, evacuationmodeling, eventually leading to technical requirements formonitoring stations and recommendations for adaptation andmitigation strategies. In large part, scientific studies to dategive insight into a single specific part of the early warningprocess.

Assessing historic or potential future tsunami impacts tocoastal communities is presented in the literature over thelast few decades. A broad overview of the achievements intsunami science is given in Synolakis and Bernhard (2006).A variety of numerical tools were developed that are capableof simulating the entire process of tsunami generation, prop-agation and run-up or parts of this process on either rect-angular grids (e.g. MOST in Titov and Synolakis (1998),TUNAMI-N2 in Goto et al. (1997) or unstructured grids(e.g. ANUGA in Nielsen et al., 2005) solving the Nonlinearshallow water wave equations. Higher accuracy was gainedwhen numerical models were developed solving Boussinesq-type equations but with much greater computational cost(e.g. Lynett et al., 2002). A tsunami inundation study dealingwith western Sumatra and our focus area presents inundationmaps based on historic events and plausible future scenariosdeduced from GPS measurements (Borrero et al., 2006) butonly on coarse available geometry. Burbridge et al. (2008)present a probabilistic tsunami hazard assessment for scenar-ios in the Indian Ocean. Regarding the devastating impactof actual tsunamis Reese et al. (2007) analyzed post-tsunamiimpact by field surveys with the intention to calibrate futuremodels for inundation depth assessment as well as for sta-bility assessment of buildings. Tinti et al. (2005) present ananalysis based on eye-witnessed tsunami reports to give in-sight into the hazard process and impact.

In the field of tsunami vulnerability assessment studies tar-get different aspects of vulnerability. Leone et al. (2007) fo-cus on physical vulnerability correlating physical parametersof buildings visible from space with damage grades surveyedafter the 2004 tsunami in Banda Aceh, Indonesia. Papath-oma et al. (2003) and Taubenbock et al. (2008a) incorporatemultiple vulnerability indicators from the natural and builtenvironment to socio-demographics. With the focus on vul-nerable conditions of people a global survey of capacities andgaps for early warning systems, identified warning dissemi-nation and preparedness to act (response capability) as theweakest elements (ISDR Newsletter, 2006). Schneiderbauerand Ehrlich (2006) discriminate “social levels” of vulnera-bility at sub-national scale. Measuring socio-economic vul-nerability to tsunami in the context of early warning meansidentifying relative differences of spatial and social patternswith regard to access to warning information and complianceof tsunami exposed social groups (Birkmann, 2006). In anactual tsunami warning authorities face various uncertaintiesof the success of an issued warning: How many people will

respond, how fast will they act, what will they do and wherewill they go (Sorensen, 2000)? The development of evac-uation simulations relies strongly on results obtained in thefield of transportation modeling. Like in transportation, onecan distinguish static approaches, e.g., Sheffi (1985), and dy-namic approaches, e.g., Peeta and Ziliaskopoulos (2001). Atypical static evacuation simulation is MASSVAC (Hobeikaand Kim, 1998). The obvious shortcoming of static modelsis that they do not capture dynamic effects, which are highlyrelevant in evacuation situations. Consequently, many dy-namic traffic assignment (DTA) models have been appliedfor evacuation simulations, e.g., MITSIM (Jha, Moore andPashaie, 2004), DYNASMART (Kwon and Pitt, 2005), andPARAMICS (Chen and Zhan, 2004). Another aspect accord-ing to which transportation models may be classified is theirgranularity: Microscopic models represent every trip-makerindividually, whereas macroscopic models aggregate trafficinto continuous streams. All of the above DTA packagesrely on microscopic traffic models. Further microscopic ap-proaches that have been applied to the simulation of evacu-ation dynamics are cellular automata (Klupfel et al., 2003)and the social force model (Helbing et al., 2002). Randomutility models are also applicable to the microscopic model-ing of pedestrian dynamics (Bierlaire et al., 2003); however,they are yet to be applied in evacuation scenarios.

In comparison to most studies our work focuses on the de-velopment of a tsunami early warning information system athigh-end geometric and thematic detail. We aim to analyzeand identify risk based on the development of hazard scenar-ios and the assessment of vulnerability of the current systemon individual building level. Furthermore, our study tries toincorporate various scientific research fields – engineering,remote sensing and social sciences – to cover manifold per-spectives for tsunami early warning. We combine data, meth-ods and results from inundation modeling, urban morphol-ogy analysis, population assessment, socio-economic analy-sis of the population and evacuation modeling. We take theinundation dynamics into account by simulating various haz-ard scenarios; we measure and quantify the level of expo-sure; we assess the physical-technical susceptibility and thesocio-economic vulnerability of the population; we analyzeemerging interactions between the hazard, the vulnerable el-ements (people, critical infrastructure, etc.) and model dy-namic evacuation scenarios with the objective to identify bot-tlenecks of the situation. The results of this interdisciplinaryresearch serve as a basis for recommendations and identifi-cation of related spatial planning efforts to reduce the riskof human and material losses in case of a potential tsunamiimpact.

The work presented here is embedded in the NumericalLast Mile Tsunami Early Warning and Evacuation Infor-mation System (Last-Mile) project (Birkmann et al., 2008a;www.last-mile-evacuation.de). The term “Last-Mile” is un-derstood as synonym for the immediate coastal area, wherea potential tsunami wave piles up and eventually hits the

Nat. Hazards Earth Syst. Sci., 9, 1509–1528, 2009 www.nat-hazards-earth-syst-sci.net/9/1509/2009/

www.last-mile-evacuation.de

H. Taubenbock et al.: “Last-Mile” preparation to a potential disaster 1511

vulnerable system on land. The project focuses on the ur-ban coastal hinterland of the city of Padang, West Sumatra,Indonesia, which has been indicated as one of the most plau-sible localities for a tsunami of disastrous proportions in thenear future (Borrero et al., 2006).

Padang is the capital city of the Sumatera Barat provincein West Sumatra, and with almost one million inhabitantsthe third largest city on the island of Sumatra, Indonesia(Fig. 2a). The complex and dynamic urban system of Padangfeatures a high concentration of population, infrastructureand economic values. Padang features supra-regional rele-vance with an international airport, a port as well as bind-ing to the rail network. Thus, the city possesses a centraleconomical role for the coastal region and the mountainousback-country. It is located directly on the coast and partiallysited beneath the mean sea level. Off the coast of Padang,the Sunda Arc marks an active convergent plate boundary.Thus, the city is located in a zone of extreme high probabilityto severe earthquakes and potential triggered tsunamis. Ac-cording to the tsunami catalogue of the Indian Ocean, whichincludes about ninety tsunamis, 80% of the tsunamis origi-nate in the Sunda arc region, where on average tsunamis aregenerated once in three years (Rastogi and Jaiswal, 2006).

A conceptual overview on risk and vulnerability serves asthe outline of this study (Sect. 2). After the introduction ofthe various data sets used (Sect. 3), the further organizationof the chapters is based on the conceptual overview: A haz-ard and a vulnerability part (Sect. 4). The results from thevarious scientific disciplines open up into main findings andrecommendations (Sect. 5), before we summarize our workand give an outlook on intended progress within the scope ofour study (Sect. 6).

2 Conceptual overview

Risk specifies something prospective and imaginary, and thusimplies uncertainty. In context of disaster risk, the conceptof vulnerability has evolved out of the social sciences andwas introduced as a response to the purely hazard-orientedperception (Schneiderbauer and Ehrlich, 2004) in the 1970s.There are various definitions and concepts on risk and vul-nerability from different schools and disciplines (Birkmann,2006; Cannon et al., 2003; Correll et al., 2001; Davidson,1997; IPCC, 2001; Blaikie et al., 1994; Turner et al., 2003;UN/ISDR, 2004); the BBC Framework (Birkmann, 2006)bases on these studies and is used to structure and conceptu-alize risk and vulnerability assessment within our approach(Fig. 1).

According to the BBC framework, risk is understood asthe interaction of the hazard and the vulnerability of a sys-tem exposed. Vulnerability itself is considered as a con-figuration and interaction of physical, demographic, social,economic, environmental and political exposed and suscep-tible elements as well as coping capacities of these elements

Fig. 1. BBC Framework (Source: Birkmann, 2006, based on Boga-rdi and Birkmann, 2004 and Cardona, 1999, 2001).

(e.g. social groups, infrastructures, economic sectors) (Eqs. 1and 2).

Risk = f (Hazard, Vulnerability) (1)

Vulnerability = f (Exposed Elements,

Susceptibility, Coping Capacity) (2)

The effect of a natural hazard on the objects or people ofa particular area/system exemplify the complex interrela-tionships and emerging domino effects (Taubenbock et al.,2008a). The BBC-framework outlines two potential paths forreducing disaster risk and vulnerability: 1) preventive mea-sures such as spatial planning and awareness raising beforea disaster manifests (t: 0); 2) disaster management such asevacuation and emergency response during a disaster (t: 1)(Birkmann, 2006). The intervention system – e.g. existingemergency management plans – has direct effects on theconfiguration of the overall vulnerability and determines thelevel of risk. One main advantage of this framework is theclearly indicated feedback loop system of the measures (in-tervention system) in reducing the risk. This implies that theentire risk and vulnerability assessment should take into con-sideration the existing or planned measures and that, at thesame time, the existing and planned measures should ensurevulnerability reduction.

The BBC conceptual framework stresses the fact that vul-nerability assessment should take into account exposed, sus-ceptible elements and coping capacities (we follow this struc-ture in Sect. 4). They determine the likelihood to suffer harmdue to potential tsunamis. The vulnerability assessment intsunami early warning context refers to identify the causality

www.nat-hazards-earth-syst-sci.net/9/1509/2009/ Nat. Hazards Earth Syst. Sci., 9, 1509–1528, 2009


of ineffective responses to early warning generating loss oflives (and livelihoods) during an emergency situation. Ac-cording to UN/ISDR (2006) this conforms to the weakestelements of early warning systems “warning disseminationand preparedness to act”.

Based on this conceptual framework we address in ourstudy the following specific questions about tsunami riskcomponents to cover required information for appropriaterecommendations and decision-making at the “last-mile”:

– What kind of inundation dynamics and what timeframes are expected?

– How many structures and people and which groups ofpeople are exposed for particular hazard scenarios?

– What factors influence different responses/behaviour tothe tsunami early warning and coping capacity of vari-ous social groups in emergency situation?

– Which bottlenecks arise during the processof evacuation?

– Which scenario-specific decisions and recommenda-tions have to be met to minimize disaster risk?

3 Multidisciplinary requirements for data sets

Problems associated with hazard analysis, risk and vulner-ability assessment, and developing risk reduction measuresare inherently spatial in nature. High resolution and updateddata on topography, bathymetry or thematic maps of the cityof Padang were not available or existing. Therefore, variousremotely sensed and hydrographical data sets were acquiredas well as field surveys containing household questionnaireswere conducted to remedy the gap.

3.1 Remotely sensed data

We use high resolution satellite imagery from the Ikonos sen-sor with geometric capabilities of 1 m in panchromatic and4 m in multispectral (blue, green, red and nir) bands to ex-tract highly detailed information on the individual physicalobjects of complex urban areas – buildings, streets and openspaces. In addition an aerial survey was conducted in Padanggathering appropriate data with even a higher resolution us-ing the Multi Functional Camera (MFC). The unique MFCsystem is one of the most advanced digital aerial line scannersystems worldwide. Characteristics of the camera includesynchronous image data acquisition, with up to 16 cm spatialresolution, and detailed and highly automated processing ofDigital Surface Models (DSM). The MFC system is one offew systems available today that is capable of producing im-age and elevation data with the appropriate radiometric andgeometric quality for this kind of project.

Generation of a homogeneous map requires post-processing of all image data. The Ikonos data were pan-sharpened and were subject to atmospheric corrections utiliz-ing the ATCOR (Atmospheric and Topographic Correction)Software (Richter, 1996). The airborne data were processedusing internal software, developed by the DLR to add naviga-bility, as recorded by GPS and the Inertial Measurement Unit(Borner et al., 2008). The resulting data were true orthopho-tos with a resolution of 25 cm. The data were mapped anddelivered in units of UTM/WGS 84. The aerial survey wasaccompanied by a building facade survey. Facades of sam-ple sections of the town were photographed and their posi-tions measured by GPS. A database of buildings was permitsidentification and visualization of evacuation and infrastruc-ture relevant buildings in the 3-D city model.

After the successful aerial survey mission, the raw datahad to be further processed, including the radiometric en-hancement and mosaicking of all aerial transects. The de-rived surface model with a horizontal as well as vertical res-olution of 25 cm required some manual topographic correc-tions such as break lines along rivers and coastal areas, whichare necessary to ensure the topological correctness of the de-rived Digital Elevation Model (DEM) that forms the basis forthe hydrodynamic calculations under consideration (Moderet al., 2008). Finally, a cross-check was conducted to ensurequality of the topographic datasets by means of fast-staticGPS measurements at twelve randomly chosen coordinatesin Padang. The accuracy resulting from this procedure isin the range of 10 cm for the horizontal and 20 cm for thevertical dimension. In addition, an interferometric digital el-evation model (DEM) based on X- and C-band data fromthe Shuttle Radar Topography Mission (SRTM) acquired inFebruary 2000 was available to analyze the terrain situation.It is a surface model with a pixel-spacing of 25 m and a ver-tical accuracy of 4–6 m (Marschalk et al., 2004).

3.2 Hydrographical survey data

The main objective of the conducted hydrographical multi-beam survey off the coast of Padang was to geometrically en-hance the existing set of bathymetrical data originating fromformer singlebeam surveys and nautical charts. Thereforewe equipped a local fishing boat with a SeaBeam 1185 echosounder (L3 Communications Elac Nautik GmbH) to a cus-tomized bracket at the hull. An inertial measurement unit(F180, CodaOctopus) and a global positioning system (Trim-ble 5700 receiver) were mounted additionally. Through-out the ship cruise profiles of conductivity, temperature anddepth were taken by a CTD probe along the vertical to adjustthe sound velocity for the sonar system.

The horizontal positioning was optimized in post-processing by means of differential GPS using correctiondata from the PSKI base station on Sikuai island. Correctionfor tidal influences was accomplished by application of timeseries derived from tide station at Teluk Bayur harbour. The



3-dimensional view showing bathymetry and topography from -50 m (blue) to 0 m (green) to 10 m (brown) and higher (white)

c) Detailed view on urban objects from the MFC

1000 surveyed buildings

0 0.5 1.0 1.5 2.0km

I n d o n e s i a Padang

Jakarta

Kuala Lumpur

Singapur

a) Overview

I N D O N E S I A

Ikonos imagery

0 0.5 1.0 1.5 2.0km

1000 surveyed buildings

Center Padang

b)

d)

Urban hinterland

2c

Airport

Harbour

Fig. 2. Overview on the location of the study area Padang, Indonesia and presentation of the acquired data sets.

area covered by the conducted multibeam survey comprisesapproximately 42 km2. Water depth varies from –3.14 m to–50.65 m in the area of interest. The post-processing of thesurveyed bathymetry was done with the software “Caris Hips& Sips”. As a result a surface model of the surveyed areawith a pixel-spacing of 3 m was produced. All water depthsare referred to mean sea level.

Even if an overall quality and error estimation for the inte-grated measurement system is difficult, the accuracy of singlecomponents could be specified instead. The inertial measure-ment unit has an attitude accuracy of 0.061◦ and a headingaccuracy of 0.175◦. The accuracy of the post-processed GPSdata is in the range of 5.0–7.0 cm whereas the sound velocitycalculation by the CTD probe is accurate to 0.2 m/s. The ac-curacy of the echo sounder device is specified as>0.2% ofthe sounding depth.

3.3 Socio-economic data

Existing statistical data only provide limited information onsocioeconomic parameters of the population. From the cen-sus in 2000 no specific information on early warning anddisaster preparedness at the local level/sub-district level areavailable. For up-to-date knowledge on socio-economic pa-rameters a survey was conducted. The needed informa-tion was primarily collected by means of geocoded house-hold questionnaires with a sample size of 1000 householdsin Padang (Setiadi, 2008; Birkmann et al., 2008b). The dis-tribution of the questionnaire samples is selected based onthe physical analysis of the urban morphology (Sect. 4.2.1),function of the building, the geographic location and ex-ternal socio-economic information on district level to coverthe entire spatial and thematic spectrum of the urban land-scape (Fig. 2d). The questionnaires encompass ques-tions on the general socio-demographic and socio-economic



characteristics of the households such as age, main and dailyactivities, income, education, and on specific themes relatedto knowledge and experience on tsunami warning and evac-uation, access to warning media, up to risk perception andintended behaviour. Relationships between parameters un-der respective themes are analyzed using both descriptivestatistics and multivariate analysis methods. This analysistakes into account the spatial parameters that contribute tothe people’s perception and capability to conduct evacuation,e.g. distance of the household to the coast and accessibilityof safe places. In order to validate the analysis results, qual-itative on field data collection at the later phase of the studyis planned.

Figure 2 displays a multi-scale overview for the studyarea Padang and acquired data sets. It shows the locationof Padang in Southeast Asia (Fig. 2a). It also shows thehighly detailed bathymetry and the digital elevation model(Fig. 2b). The detailed view of the MFC data displays thehighly detailed geometric capabilities of the sensor (Fig. 2c).The Ikonos data provide an overview on the heterogeneous,small-scale and irregular urban pattern characterized by thefundamental physical elements of buildings, streets and openspaces in the entire urban area of Padang and indicates loca-tions of the center or transportation hot spots. It also displaysthe spatial distribution of the sampling households, whichwas intended to cover the complete spectrum of the urbansystem – from buildings in the urban centre to the suburbs,from the coast to the urban hinterland and from small andlow structures to large and high-rise buildings (Fig. 2d).

4 Interdisciplinary information system

This chapter specifies methods and results from the variousscientific fields – inundation modeling, urban morphologyanalysis, population assessment, socio-economic analysis ofthe population and evacuation modeling – and their inter-linkage. The aim is to highlight different components of theconceptual framework (Fig. 1) to substantially contribute totsunami early warning and an evacuation information sys-tem. Knowledge on the probability, severity and the spatialimpact of the hazard – in our case a tsunami – is the first ele-ment for an intermeshing information system. The second el-ement is the vulnerability of a system, in our case the coastalurban area of Padang, with its exposed elements like struc-tures, population or economy, with its susceptibility and withits coping capacity. Interdisciplinary collaboration aims tohighlight and analyze dependencies within the process chainto link the various different scientific perspectives for a moreholistic understanding of disaster risk and its management.

4.1 Hazard

A tsunami is a set of ocean waves caused by any large, abruptdisturbance of the sea surface (Bernhard et al., 2006). These

by Alfred-Wegener-Institute (AWI) based on the model TsunAWI (Harig et al, 2007). The NLSW equations are employed as a basic set of equations that are solved on an unstructured finite-elements mesh (Behrens, 2008). Subsequently a hydrodynamic inundation modeling tool called ANUGA (Australian National University and Geoscience Australia, Nielsen et al. 2005) is employed for the computations of run-up and on-land flow for the project area due to effective shock capturing methods and domain decomposition strategies. The interface between both numerical models is designed to linearly interpolate model outputs from TsunAWI onto the mesh nodes at the ocean boundary of the ANUGA domain (Fig 3a). We have chosen to simulate the inundation of the city of Padang in a hybrid model mainly due to computational cost and modeling time considerations. Additionally ANUGA assumes that all spatial coordinates are within one UTM zone. Modeling the end-to-end process of generation and propagation of tsunami as well as inundation in a single model is therefore currently inapplicable for the Padang area by ANUGA software since the earthquake source and the area of interest are not always in the same UTM zone.

ANUGA uses a finite-volume approach to solve the nonlinear shallow water wave equations in conservative form. A spatially varying triangular mesh is used to discretize the considered area of interest. The conserved quantities are water depth and horizontal momentum in x and y direction (Nielsen et al. 2005). A detailed description of the applied numerical scheme can be found in the work of Zoppou & Roberts (1999). The chosen coupling strategy only allows for unidirectional propagation of the incident wave into the ANUGA domain. Care must therefore be taken when the domain is established since both models run independently from each other and don't interact or exchange results while simulations are being executed.

A twofold strategy is adopted throughout our modeling for the city of Padang in order to satisfy the fact that the urban topology of our focus area is densely built-up. On the one hand we employ two different house masks derived from the remotely sensed imagery. The finest house mask bases on the real building

Fig. 4. General idea behind coupling the two different numerical models, light green line indicates the interface between AWI and ANUGA model, red dots represent AWI mesh nodes, conserved quantities are linearly interpolated to the ANUGA boundary, and rectangular box represents parts of the ANUGA domain

a)

b)

Fig. 3. a) Locations of epicentres for the pre-calculated scenarios by AWI available with Mw 8.5 and Mw 9.0; b) Overview of the modelled area, Padang city, Teluk bayur and the oil harbour, the wave is introduced into the model at the western boundary

Boundary of numerical model

Fig. 3. (a)Locations of epicentres for the pre-calculated scenariosby AWI available with Mw=8.5 and Mw=9.0; (b) Overview of themodelled area, Padang city, Teluk bayur and the oil harbour, thewave is introduced into the model at the western boundary.

disastrous waves could possibly be triggered by underseaearthquakes, land slides or volcano ruptures. To assess theimpact of tsunamis, in-depth studies of inundation dynamicsextracted from hydrodynamic simulations could contributewidely, where detailed information about velocity fields andflow depth is needed for evacuation simulation in coastal ur-ban areas. From the field of coastal engineering these pa-rameters could be obtained by solving the appropriate equa-tions describing the predominant character of the flow field.Tsunamis are referred to as surface gravity waves with a typ-ical wave length that is several times the water depth. Thiswave phenomenon is often represented mathematically bythe nonlinear shallow water (NLSW) equation or Boussi-nesq type equations (Pedersen et al., 2008). We have chosenNSLW for the hydrodynamic study of our project area due tocomputational costs and since dispersive effects in the nearshore zone are less dominant.

The computational domain comprises an area of about580 km2 that is bordered on the southern, western and north-ern part by the Indian Ocean. In the east of the domain the



by Alfred-Wegener-Institute (AWI) based on the model TsunAWI (Harig et al, 2007). The NLSW equations are employed as a basic set of equations that are solved on an unstructured finite-elements mesh (Behrens, 2008). Subsequently a hydrodynamic inundation modeling tool called ANUGA (Australian National University and Geoscience Australia, Nielsen et al. 2005) is employed for the computations of run-up and on-land flow for the project area due to effective shock capturing methods and domain decomposition strategies. The interface between both numerical models is designed to linearly interpolate model outputs from TsunAWI onto the mesh nodes at the ocean boundary of the ANUGA domain (Fig 3a). We have chosen to simulate the inundation of the city of Padang in a hybrid model mainly due to computational cost and modeling time considerations. Additionally ANUGA assumes that all spatial coordinates are within one UTM zone. Modeling the end-to-end process of generation and propagation of tsunami as well as inundation in a single model is therefore currently inapplicable for the Padang area by ANUGA software since the earthquake source and the area of interest are not always in the same UTM zone.

ANUGA uses a finite-volume approach to solve the nonlinear shallow water wave equations in conservative form. A spatially varying triangular mesh is used to discretize the considered area of interest. The conserved quantities are water depth and horizontal momentum in x and y direction (Nielsen et al. 2005). A detailed description of the applied numerical scheme can be found in the work of Zoppou & Roberts (1999). The chosen coupling strategy only allows for unidirectional propagation of the incident wave into the ANUGA domain. Care must therefore be taken when the domain is established since both models run independently from each other and don't interact or exchange results while simulations are being executed.

A twofold strategy is adopted throughout our modeling for the city of Padang in order to satisfy the fact that the urban topology of our focus area is densely built-up. On the one hand we employ two different house masks derived from the remotely sensed imagery. The finest house mask bases on the real building

Fig. 4. General idea behind coupling the two different numerical models, light green line indicates the interface between AWI and ANUGA model, red dots represent AWI mesh nodes, conserved quantities are linearly interpolated to the ANUGA boundary, and rectangular box represents parts of the ANUGA domain

a)

b)

Fig. 3. a) Locations of epicentres for the pre-calculated scenarios by AWI available with Mw 8.5 and Mw 9.0; b) Overview of the modelled area, Padang city, Teluk bayur and the oil harbour, the wave is introduced into the model at the western boundary

Boundary of numerical model

Fig. 4. General idea behind coupling the two different numericalmodels, light green line indicates the interface between AWI andANUGA model, red dots represent AWI mesh nodes, conservedquantities are linearly interpolated to the ANUGA boundary, andrectangular box represents parts of the ANUGA domain.

city of Padang is introduced taking into account the highly-resolved digital elevation and surface models surveyed by anairborne remote sensing campaign (cp. Sect. 3.1). Besidesthe main part of Padang we include the economically fun-damental harbour regions Teluk Bayur and the oil harbourin the south of Bungus Bay. In addition we add a strip of3km width in the north made up of less accurate elevationdata available from the Intermap campaign (Intermap, 2007).The landward boundary is confined by the 80 m contour line.A great demand for highly-resolved bathymetry in the near-shore zone and topography for adequately modeling over-land flow was stated e.g. by McCloskey et al. (2008). Small-scale bathymetry variations affect modeling results to a ma-jor order (Borrero et al., 1997) and possible time-dependentchanges in coastal topography need to be considered whennumerical models are validated (Borrero et al., 2006). Cop-ing with these constraints we conducted a hydrographicalsurvey off the coast of Padang in 2007. This 3×3 m gridobtained from that campaign is merged into a final digitalelevation model together with the newly surveyed topogra-phy (cp. Sect. 3.1). Amongst these highly-resolved data setswe use SRTM data, GPS measurement, digitised nauticalcharts, General Bathymetric Chart of the Oceans (GEBCO)and C-Map bathymetry (Jeppessen Marine) products in or-der to finalize the geometrical basis for the numerical mod-eling. Given the fact that we used the most accurate, highlyresolved and contemporary datasets for bathymetry and to-pography, time-dependent changes have been considered inthe geometry model sufficiently. Other time-dependent ef-fects due to sediment transport or debris flow are beyond thescope of this study.

For the tsunami inundation modeling a hybrid approachwas chosen to study wave propagation from the source ontodry land. In a basic step the source modeling and thedeep-water propagation was carried out by Alfred-Wegener-Institute (AWI) based on the model TsunAWI (Harig et al.,2007). The NLSW equations are employed as a basic set ofequations that are solved on an unstructured finite-elementsmesh (Behrens, 2008). Subsequently a hydrodynamic inun-

dation modeling tool called ANUGA (Australian NationalUniversity and Geoscience Australia, Nielsen et al., 2005)is employed for the computations of run-up and on-land flowfor the project area due to effective shock capturing meth-ods and domain decomposition strategies. The interface be-tween both numerical models is designed to linearly interpo-late model outputs from TsunAWI onto the mesh nodes at theocean boundary of the ANUGA domain (Fig. 3a). We havechosen to simulate the inundation of the city of Padang in ahybrid model mainly due to computational cost and modelingtime considerations. Additionally ANUGA assumes that allspatial coordinates are within one UTM zone. Modeling theend-to-end process of generation and propagation of tsunamias well as inundation in a single model is therefore currentlyinapplicable for the Padang area by ANUGA software sincethe earthquake source and the area of interest are not alwaysin the same UTM zone.

ANUGA uses a finite-volume approach to solve the non-linear shallow water wave equations in conservative form.A spatially varying triangular mesh is used to discretize theconsidered area of interest. The conserved quantities are wa-ter depth and horizontal momentum in x- and y-direction(Nielsen et al., 2005). A detailed description of the appliednumerical scheme can be found in the work of Zoppou andRoberts (1999). The chosen coupling strategy only allowsfor unidirectional propagation of the incident wave into theANUGA domain. Care must therefore be taken when thedomain is established since both models run independentlyfrom each other and don’t interact or exchange results whilesimulations are being executed.

A twofold strategy is adopted throughout our modeling forthe city of Padang in order to satisfy the fact that the urbantopology of our focus area is densely built-up. On the onehand we employ two different house masks derived from theremotely sensed imagery. The finest house mask bases onthe real building situation in the city including size, form andheight of every individual house (Sect. 4.2.1). In addition, weuse a coarse house mask combining the individual buildingsper unit (blocks which are defined by the street network) toone quadrangle with a size corresponding to the real size ofall individual buildings summed up. A major reason for run-ning models with both DSM and DEM plus house masks isthat it presently is not fully understood how macro-roughnesselements could be taken into account in these kinds of hy-drodynamic studies. The coarse house mask reduces the pro-cessing time significantly. On the other hand we employ thehighly-resolved digital surface model that is sufficiently rep-resented by a maximal triangle area of about 20 m2. Themaximal triangle area in the triangular mesh used is a mea-sure for the quality of the hydrodynamic study. The valueschosen are optimized on the one hand to capture all relevanttopographic features and on the other hand to reduce needlesscomputation time. As evacuation routes and upgrading rec-ommendations are a focal point in our interdisciplinary study,the resulting flow fields need to represent at least major roads,



(a) (b)

(c) (d)

Fig. 5. Sequential snapshots of combined inundation extend map and inundation depth map for Padang City, times are measured with respectto the arrival of the leading wave at the shoreline at zero seconds. Upper left: Sequence at 60 s, upper right: Sequence at 240 s, lower left:Sequence at 420 s and lower right: Sequence at 600 s.

streets and waterways in the city to compare with evacuationsimulation adequately (cp. Sect. 4.2). This requirement isaddressed by using meshes that are in the range of 10–20 m2

for the city of Padang.

A common approach to calibrate and validate numericalstudies is to compare them with historic tsunami inundationpatterns, but for the city of Padang, the availability of vali-dation events for our numerical studies is very poor (Borreroet al., 2006). Therefore we decided to rely on a large setof pre-calculated scenarios with spatially distributed epicen-tres (Fig. 3a). From the considered scenarios the mediumwarning time measured from the earthquake event until thearrival of the leading wave front is approximately 40 min,minimum 23 min and maximum 65 min (oral communicationwith AWI). Based on these scenarios hazard zones are iden-

tified and communicated to conduct adjacent hazard and riskassessment studies (cp. Sect. 4.2). The pre-calculated sce-narios were taken as boundary conditions for both a prelim-inary study as well as for the full model domain. The resultfrom the preliminary study revealed some important hints onthe further modeling published in Goseberg et al. (2009) andGoseberg and Schlurmann (2008). The most striking fact isthat the employments of highly-resolved data sets in contrastto commonly available data lead to channelized flow in ma-jor streets together with much higher currents. When a moresophisticated DEM is used not only flow velocities are am-plified, but also flow depths and water levels directly at theshoreline are increased.

Finally these conclusions from the preliminary study couldbe confirmed and deepened as we analyzed first results from



Fig. 6. Exemplary flood extent maps for the different modeling strategies applied. Left: Maximum inundation depth [m] for a model withdigital surface model; Right: Maximum inundation depth [m] for a model using a medium resolution house mask derived from remotelysensed data.

Fig. 7. Exemplary hazard maps from preliminary results of our multi-scenario approach compiled from 21 scenarios with underlying DSMmodel, counts are normalized by the maximal number of scenarios, left: city center of Padang, right: harbour Teluk Bayur south of the city.

the main study of the domain. In this paper we select onlytwo different tsunami sources to show results with respect tothe interdisciplinary approach. For details of tsunami sourceswe refer to Harig et al. (2007). Figure 5 presents sequentialsnapshots of inundation dynamics for a given Mw=9.0 ex-emplary event south-east of Pulau Siberut for the model runswith the underlying digital elevation model and house masks.These maps combine inundation extent and spatial informa-tion about water depth for the given point of times with re-spect to the arrival of the leading wave at the shoreline at zeroseconds. It becomes apparent that areas near major channelsand rivers draining the coastal plain are affected earlier thanareas more distant to the water bodies. Canalised flow pattern

are generally observed when house masks are used instead ofdigital surface models. As a matter of fact, this flow patternresults in higher water levels near the coast and higher ve-locities where house polygons are absent in the domain. Onthe contrary, more distant areas to the shore face significantlylower water levels compared to model runs with DSM.

Figure 6 shows maximum inundation extents for two dif-ferent model runs for the two modeling strategies discussedabove. On the left hand side maximum inundation is plot-ted using the digital surface model of Padang. Due to themesh resolution open spaces and built-up areas are not dis-tinguishable from the result, since buildings that are clearlyrepresented in the DSM degenerate under the applied mesh



interpolation algorithms. Nevertheless, building structuresfrom the DSM affect model results to a significant level com-pared to plain DEM that are commonly used for such simu-lations. We found that using DSM data or DEM data to-gether with a house mask for our numerical studies, inun-dation extent is less compared to simulation runs using onlyDEM data. That is due to the fact that flow is impeded bythe existents house features in the model. The overland flowis reflected or decelerated while interacting with or being re-flected at houses. On the other hand maximum inundationplotted for a simulation combined with a medium resolvedhouse mask influences inundation dynamics as well.

Finally we deduced probabilistic hazard maps from themodeling approach discussed above. The hazard maps pre-sented in Fig. 7 are compiled from the scenarios already postprocessed (Mw=9.0) by means of counting any inundatedraster cell in each scenario. For the probabilistic analyzesinundated raster cells get a value of 1 and dry raster cells avalue of 0. The resulting hazard map is deduced by summingup the available scenarios. Figure 7 shows normalized val-ues with respect to the maximal number of tsunami scenariosin the same moment magnitude classified in 4 hazard zones.The resulting map products are finally being delivered to as-sess perspectives of vulnerability and to improve evacuationplanning.

4.2 Vulnerability

With respect to our conceptual guideline – the BBC frame-work – risk is the interaction of the hazard (cp. Sect. 4.1)and the vulnerability of a system exposed. In the followingsection we attempt to measure and analyze vulnerability con-sidered as a configuration and the interaction of exposed ele-ments and their susceptibility as well as coping capacities ofthese elements (e.g. social groups, infrastructures, economicsectors).

4.2.1 Exposed elements

Exposure of elements is an inherently spatial problem. Re-mote sensing is a suitable scientific field to provide area-wide, up-to-date and accurate information on the urban landcover and its spatial structural pattern – the exposed ele-ments. Based on the Ikonos imagery (Fig. 2d) we use a com-bination of an automatic object-oriented, hierarchical clas-sification methodology (Taubenbock and Roth, 2007; Kasset al., 2007) and subsequent manual enhancement to derivea thematic land cover classification. The automatic clas-sification approach resulted in 86.1% accuracy. Problemsare automatic delineation of individual buildings in highlydense built-up areas as well as the automatic differentiationbetween certain roof types and sealed areas, which madethe manual enhancement necessary. The results are eightclasses mapping the urban structure – houses, streets, sealedareas, grassland, trees, wetland, bare soil, and water. No

ground truth data were at hand for an accuracy assessment.Thus accuracy has been analyzed by a randomization of500 checkpoints and a subsequent visual verification processbetween the classification and the Ikonos imagery calculat-ing an error matrix using producer and consumer accuracies.The overall accuracies resulted in 97% correct classified pix-els.

Value-adding from 2-D land-cover information to a 3-D city model has been performed. Using the digital sur-face model in combination with the building mask we assessthe height of individual structures. In addition we calculatebuilding sizes, roof types, and built-up density (Taubenbocket al., 2008b). Utilizing the field work experience in combi-nation with the physical parameters of the individual build-ings – size, height, location, alignment and roof type – landuse is assessed as an additional feature of every building inthe entire urban landscape of Padang. This enables to dif-ferentiate between residential, mixed, commercial and indus-trial usage on a basic level. Thus, the city model even enablesa basic localization of economic focal areas. A higher the-matic specification is integrated using field work informationto localize critical infrastructure like hospitals or schools.The orographic condition of the urban region of Padang uti-lizing the digital elevation model completes the urban envi-ronment of the coastal city. Thus, a three-dimensional citymodel displays the complex, heterogeneous and fine-scaleurban morphology and urban land use in a very detailedmanner. This product displays essential basic informationto know “what” is “where” in the city of Padang. It alsoprovides the spatial basis for interdisciplinary research: Thebuilding mask is used as input parameter for inundation mod-eling of a potential tsunami (Sect. 4.1) and the derived streetnetwork is used for evacuation modeling (Sect. 4.2.2). Fur-thermore the 3-D city model is used to systematically dis-tribute the survey samples (Sect. 4.2.2). Multi-layer analysisallows quantifying and localizing exposed elements of theurban landscape. Figure 8 shows a perspective view from thesouth on the city model of Padang.

Static urban morphology characterizes an urban agglom-eration. But the dynamics of its population also distinguishthe urban exposure. We correlate the static element of physi-cal urban structures with the dynamic element of spatial andtemporal behaviour of the population by means of the citymodel and surveyed data on punctual population informa-tion. Utilizing the building size and its particular height us-able building spaces (living spaces) are calculated. Thus, cal-culation of average number of inhabitants per square meteris derived based on 500 surveyed geocoded sample build-ings distributed around Padang (Taubenbock et al., 2008c).The mathematical concept of bottom-up extrapolation hasbeen presented in detail by Taubenbock et al. (2007). A lin-ear extrapolation algorithm projects the people proportion-ally to the living spaces of the particular house. Accuracyassessment has been carried out resulting in 86.73% com-pliance. In addition the detailed knowledge on the usage



buildings - size, height, location, alignment and roof type - land use was assessed as an additional feature of every building in the entire urban landscape of Padang. This enables to differentiate between residential, mixed, commercial and industrial usage on a basic level. Thus, the city model even enables a basic localization of economic focal areas. A higher thematic specification was integrated using field

Fig. 8. City model of Padang derived from high resolution satellite data and a digital elevation model (Intermap, 2007)

work information to localize critical infrastructure like hospitals or schools. The orographic condition of the urban region of Padang utilizing the digital elevation model completes the urban environment of the coastal city. Thus, a three-dimensional city model displays the complex, heterogeneous and fine-scale urban morphology and urban land use in a very detailed manner. This product displays essential basic information to know ‘what’ is ‘where’ in the city of Padang. It also provides the spatial basis for interdisciplinary research: The building mask is used as input parameter for inundation modeling of a potential tsunami (4.1.) and the derived street network is used for evacuation modeling (4.2.2). Furthermore the 3-D city model is used distribute the survey samples (4.2.2). Multi-layer analysis allows quantifying and localizing exposed elements of the urban landscape. Figure 8 shows a perspective view from the south on the city model of Padang. Static urban morphology characterizes an urban agglomeration. But the dynamics of its population also distinguish the urban exposure. We correlate the static element of physical urban structures with the dynamic element of spatial and temporal behaviour of the population by means of the city model and surveyed data on punctual population information. Utilizing the building size and its particular height usable building spaces (living spaces) are calculated. Thus, calculation of average number of inhabitants per square meter is derived based on 500 surveyed geocoded sample buildings distributed around Padang (Taubenböck et al 2008c). The mathematical concept of bottom-up extrapolation has been presented in detail by Taubenböck et al. (2007). A linear extrapolation algorithm projects the people proportionally to the living spaces of the particular house. Accuracy assessment has been carried out resulting in 86.73 % compliance. In addition the detailed knowledge on the usage of the buildings is at hand. Thus, the bottom-up extrapolation algorithm is calibrated on the different utilization of the buildings –residentially, mixed or commercially used – with respect to the time of day. From the questionnaire, day- and night-time population for the sample buildings are at hand, which is included to assess on average the quantity of people per building with respect to usage. The result is a highly detailed knowledge on the dynamic spatial behaviour and time-dependent whereabouts of people within the complex urban landscape (Fig. 9). Thus, a spatial shift in

Legend Buildings – land use residential mixed usage / harbour hotel market / shopping mall industry commercial use / offices university / school mosque hospital

Land cover streets sealed areas grassland forest bare soil water

Fig. 8. City model of Padang derived from high resolution satellite data and a digital elevation model (Intermap, 2007).

Fig. 9. Population distribution on building level.

of the buildings is at hand. Thus, the bottom-up extrapola-tion algorithm is calibrated on the different utilization of thebuildings – residentially, mixed or commercially used – withrespect to the time of day. From the questionnaire, day- andnight-time population for the sample buildings are at hand,which is included to assess on average the quantity of peopleper building with respect to usage. The result is a highlydetailed knowledge on the dynamic spatial behaviour andtime-dependent whereabouts of people within the complexurban landscape (Fig. 9). Thus, a spatial shift in exposure ismapped that influences location and quantities of potentiallyaffected people in dependency of the time of day.

Table 1. Quantification of affected buildings and time-dependentassessment of affected people with respect to the hazard map(Fig. 7).

Name of Probability Buildings Buildings People affectedzone affected affected Day-time Night-time

No risk 0.0% 50.657 57.56% 285 600 281 500Low risk 0.01–5.0% 4.213 4.79% 36 900 32 400Medium risk 5.01–15.0% 17.418 19.79% 98 800 102 100High risk 15.01–45.0% 14.308 16.26% 81 900 75 700Extreme risk 45.01–100.0% 1.408 1.60% 9500 7200Total 88.004 100% 512 700 498 900

Interdisciplinary combination of the hazard map, the 3-D city model as well as the time-dependent population distri-bution allows now the probabilistic and quantitative assess-ment of scenario-dependent affected buildings and people.Table 1 shows the results for the 4 risk zones and revealsfor the worst case scenario in our test area (Fig. 2d) thatoverall 37 347 buildings would be inundated in different riskzones and 227 100 people at daytime and 217 400 at night-time would be at risk.

4.2.2 Susceptibility and coping capacity

Susceptibility and coping capacity as components of vul-nerability are defined as the degree to which a systemis able or unable to cope with the adverse effects of ahazardous impact (adapted from McCarthy et al., 2001).We attempt to measure the degree of vulnerability withinthe specific context of tsunami early warning and evacua-tion: i.e. the socio-economic vulnerability assessment shouldprovide hints on the challenges that may appear due to



Fig. 10.Various responses to the warning during the last “potential”tsunami experience.

perception and behaviour of the people. Social groups con-fronted with the tsunami risk in reality indicate gaps be-tween early warning technology and social response capabil-ity. They are not covered using pure engineering approachesor models that normally use simplified assumptions on thesocial aspects. It is intended to investigate parameters thatwould determine the susceptibility or coping capacity of var-ious social groups during evacuation in order to understandwhat actually lead to the gaps identified.

Initially, the socio-economic factors influencing the expo-sure of various social groups to tsunamis derived from dailyactivity pattern and mobility data are further investigatedusing data obtained from the household questionnaires onsocio-economic vulnerability. This provides complementaryinformation to validate the time-specific population distri-bution according to the physical structural pattern extractedfrom remotely sensed data (cp. Sect. 4.2.1).

Subsequently, further analysis on the evacuation behaviouris conducted. As a starting point the following hypotheseswere used with regard to the relevant vulnerability themes:

– various social groups have different access to earlywarning,

– various social groups would respond differently to thewarning and to the given evacuation instruction,

– various social groups would have different capabilityand considerations (e.g. evacuation routes, evacuationin groups or individually) in conducting evacuation.

These hypotheses are tested using common statisticalmethods for descriptive analysis, correlation and regressionanalysis. As an example, the findings from the householdquestionnaires showed that during the September 2007 earth-quake in the region of Padang 26.5% of people did not re-ceive any tsunami early warning (Fig. 10). 51.3% received

Fig. 11. Effectiveness of various media in disseminating the warn-ing during the last experience.

formal warning through television (TV) and radio and therest (22.2%) were only informally notified. There was vary-ing effectiveness of warning dissemination through differentmedia in proportion to availability of the respective media inthe households. This may be explained by the utilization ofthese media during the time of the event. It indicates thatsome households have to take uninformed decisions becauseof receiving different warning messages from different me-dia. Results show that especially the informal notificationand the interpretation of the warning information was dif-ficult. There were only a small proportion of people evac-uating after receiving the warning during the last experience(34%), however, the intention for evacuating in case of futuretsunami early warning showed higher rates (75%) (Setiadi,2008; Birkmann et al., 2008b) (cp. Figs. 11 and 12).

Multinomial logistic regressions are employed in order toinvestigate which parameters influence the intention to evac-uate. The preliminary results identify a model of evacuationintention including parameters with some extent of signif-icant influence, namely knowledge of tsunamis, discussionabout tsunami risk in the community, perception on people’svulnerability, knowledge of evacuation places, personal pref-erence for safe places, self-efficacy in evacuation and doubton tsunami early warning information. The aggregation ofthese parameters shows the level of awareness and attitudetowards the current early warning and evacuation plan andwill be a proxy of potential response to tsunami early warn-ing. Moreover, distance to the coast and household charac-teristics such as household size and existence of elderly alsoplay a role.

In order to estimate the capability of the people to con-duct evacuation on time, the following parameters are takeninto account: particularly parameters related to demographicdistribution of vulnerable groups with limited running speed(and access to vehicle, if evacuation using vehicle would beconsidered) as well as evacuation destination, related to evac-uation knowledge, participation in evacuation drills and per-ceived safe places.



Fig. 12. Knowledge about safe places and routes of evacuation.

Such assessments reveal socio-economic parameters,which shall be used as indicators for continuous monitor-ing the effectiveness of implemented and planned measures– e.g. education programs – to reduce the vulnerability of thepeople. The assessment not solely uses questionnaires andqualitative research, but also incorporates spatial informa-tion extracted from remote sensing. This combination is anenriching approach for assessing vulnerability comprehen-sively.

For example the punctual household questionnaires havebeen investigated regarding spatial correlation to physical ur-ban morphology. Using the physical urban morphology pa-rameters from the city model (Fig. 8) a semantic classifica-tion has been performed. The idea of semantic classifica-tion aims at a first assumed interrelation between physicallyhomogeneous sectors within the complex urban morphol-ogy and the socioeconomic characteristics of people residingthere (Taubenbock et al., 2009). The combination of the area-wide available statistical physical parameters describing thebuilding stock of Padang per sector – built-up density, aver-age house size, average building height, location – enables toidentify physically homogeneous areas. The housing qualityis assumed to be higher with rising building size or heightand declining built-up density. We classify six different se-mantic classes – “slums”, “suburbs”, “low class areas” (LC),“middle class areas” (MC) and “high class areas” (HC). Forevery semantic class we assume typical physical conditions.As an example, slum areas are defined by the highest built-updensity measured within the urban environment with mostlyone storey buildings and the smallest buildings sizes.

The hypotheses that socio-economic parameters, like e.g.the income of the people, meet the assumptions made by thesemantic classification are analyzed. The results show thatthe classified slum areas as well as the classified suburb ar-eas reveal lowest income values independent from their lo-cation within the urban landscape. In addition consistent ris-ing income levels to the semantic classes “low class” and

height and declining built-up density. We classify six different semantic classes – ‘slums’, ‘suburbs’, ‘low class areas’ (LC), ‘middle class areas’ (MC) and ‘high class areas’ (HC). For every semantic class we assume typical physical conditions. As an example, slum areas are defined by the highest built-up density measured within the urban environment with mostly one storey buildings and the smallest buildings sizes.

The hypotheses that socio-economic parameters, like e. g. the income of the people, meet the assumptions made by the semantic classification are analyzed. The results show that the classified slum areas as well as the classified suburb areas reveal lowest income values independent from their location within the urban landscape. In addition consistent rising income levels to the semantic classes ‘low class’ and subsequently to ‘middle class’ areas is found. In the southern area of Padang the observation proofs what was stated in the initial hypothesis – a rising income for the ‘high class’, while in the northern area lower incomes for this class are detected. Thus concluding, the physical urban morphology basically correlates with socioeconomic parameters of the people. The combination of area-wide available remotely sensed data with punctual survey data shows interdisciplinary value-adding. The extrapolation enables to map spatial distribution of different social groups and thus reveals for our sample indicator ‘income’ a spatial pattern of economic vulnerability.

The results of the assessment are also useful as inputs for deriving evacuation scenarios and simulation models, which will be further elaborated below. However, due to the fact that vulnerability assessment in the early warning context deals much with complex perception and behaviour questions, it is a great challenge to generate precise prediction on the evacuation behaviour in quantitative and probabilistic terms as well as to integrate it directly as parameters into the evacuation modeling. At the first phase, the results are generated in form of relative values for comparing different areas and social groups in the city and qualitative explanation as consideration for the potential bottlenecks.

The planning for an evacuation -after a tsunami warning is triggered- is an important aspect of disaster management. The advance warning time will be at most 30 minutes. From overlapping analysis of the spatial hazard impact and assessment of dynamic population distribution approximately 230.000 people are living in the affected area (Fig. 7, Fig. 9 & Tab. 1). Therefore, a detailed analysis of aspects that could influence the evacuation process is necessary. With this analysis it should be possible to:

• Give an estimate of the evacuation time. • Detect bottlenecks that could for example emerge at bridges. • Identify highly endangered areas, where a vertical evacuation seems the only possible solution.

From these results it is planned to derive evacuation recommendations. This could include evacuation maps telling the people where to move to or recommendations where to build tsunami proof shelters. Because of the complexity of the system, an analytic solution to the evacuation problem seems impossible. Therefore a microscopic multi-agent simulation for the city with all its inhabitants has been developed. With this simulation it should be possible to obtain an estimate of the evacuation process.

There are several requirements that have to be met to get a realistic simulation. The simulation needs to reflect the socio-economic profile; this is provided by the investigation described above. Information about distribution of the population as a function of time is integrated (cp. 4.2.1.), but to date, social patterns are not yet integrated in the simulations. Another important aspect for the evacuation is information about potential flooding scenarios. There is a mechanism to integrate the results from 4.1. directly in the evacuation simulation. For the physical simulation of the evacuation procedure a detailed picture of geographic structure of the city is needed. This information has mainly been extracted from

Fig. 13. Location-based mean value correlation of semantic classes derived from remotely sensed data and the socioeconomic parameter income and their standard deviation

Fig. 13. Location-based mean value correlation of semantic classesderived from remotely sensed data and the socioeconomic parame-ter income and their standard deviation.

subsequently to “middle class” areas is found. In the south-ern area of Padang the observation proofs what was stated inthe initial hypothesis – a rising income for the “high class”,while in the northern area lower incomes for this class are de-tected. Thus concluding, the physical urban morphology ba-sically correlates with socioeconomic parameters of the peo-ple. The combination of area-wide available remotely senseddata with punctual survey data shows interdisciplinary value-adding. The extrapolation enables to map spatial distributionof different social groups and thus reveals for our sample in-dicator “income” a spatial pattern of economic vulnerability.

The results of the assessment are also useful as inputs forderiving evacuation scenarios and simulation models, whichwill be further elaborated below. However, due to the factthat vulnerability assessment in the early warning contextdeals much with complex perception and behaviour ques-tions, it is a great challenge to generate precise predictionon the evacuation behaviour in quantitative and probabilisticterms as well as to integrate it directly as parameters into theevacuation modeling. At the first phase, the results are gener-ated in form of relative values for comparing different areasand social groups in the city and qualitative explanation asconsideration for the potential bottlenecks.

The planning for an evacuation – after a tsunami warn-ing is triggered – is an important aspect of disaster manage-ment. The advance warning time will be at most 30 min.From overlapping analysis of the spatial hazard impactand assessment of dynamic population distribution approxi-mately 230 000 people are living in the affected area (Figs. 7,9 and Table 1). Therefore, a detailed analysis of aspects thatcould influence the evacuation process is necessary. With thisanalysis it should be possible to:



– Give an estimate of the evacuation time.

– Detect bottlenecks that could for example emerge atbridges.

– Identify highly endangered areas, where a vertical evac-uation seems the only possible solution.

From these results it is planned to derive evacuation recom-mendations. This could include evacuation maps telling thepeople where to move to or recommendations where to buildtsunami proof shelters. Because of the complexity of the sys-tem, an analytic solution to the evacuation problem seemsimpossible. Therefore a microscopic multi-agent simulationfor the city with all its inhabitants has been developed. Withthis simulation it should be possible to obtain an estimate ofthe evacuation process.

There are several requirements that have to be met to geta realistic simulation. The simulation needs to reflect thesocio-economic profile; this is provided by the investigationdescribed above. Information about distribution of the popu-lation as a function of time is integrated (cp. Sect. 4.2.1), butto date, social patterns are not yet integrated in the simula-tions. Another important aspect for the evacuation is infor-mation about potential flooding scenarios. There is a mech-anism to integrate the results from Sect. 4.1 directly in theevacuation simulation. For the physical simulation of theevacuation procedure a detailed picture of geographic struc-ture of the city is needed. This information has mainly beenextracted from satellite imagery by detecting all the walka-ble area in a first step and generating a network from thisafterwards (cp. Sect. 4.2.1.). The network not only includesstreets but also sidewalks and squares. But not all the neededinformation could be extracted automatically. For exampleit is almost impossible to detect obstacles like fences andthe like from satellite images. Therefore, a survey was con-ducted to take geo-encoded pictures from different placeswhere bottlenecks during evacuation could emerge. Withthese pictures we can estimate the accessibility of potentialareas of refuge. More detailed information on the input datais given by Lammel et al. (2009).

The evacuation simulation is based on the MATSim(multi-agent traffic simulation,www.matsim.org) frame-work, where each evacuee is modeled as an individual soft-ware agent. Multi-agent simulation is defined as a compu-tational model for simulating the actions and interactions ofautonomous individuals with a view to assessing their effectson the system as a whole. The underlying flow model sim-ulates the traffic based on a simple queue model where onlyfree speed, bottleneck capacities, and link storage capacitiesare taken into account. The queue simulation, albeit simple,captures the most important aspects of evacuations such asthe congestion effects of bottlenecks and the time needed toevacuate the endangered area. During simulation the agentsmake independent decisions about escape routes or when toevacuate (i.e. their individual response times). Consequently,

the evacuation for every single agent is modeled separatelyand is stored in a so-called plan containing the starting timeand the evacuation route. A snapshot of the visualization ofthe simulation is shown in Fig. 14. For more technical de-tails of the simulation framework see Lammel et al. (2009)and Lammel and Nagel (2009).

The results of the simulation reveal an estimation for theoccurrence of bottlenecks and an estimated overall evacua-tion time. A major determinant for the overall evacuationtime is the number and place of shelters and places of refuge.In the simulation, these are called sinks. Different from thenatural places of refuge like hills, the number and placesof shelter (including buildings used as shelters like concretemulti-storey buildings) can be influenced by the adminis-tration (and potentially the city’s inhabitants and especiallyowners of buildings). Therefore, a second strategy within“Last-mile” (next to providing information and guidance onevacuation strategies) is to provide recommendations on thenumber and location of shelters and to identify buildings andrecommend their use as shelters. A detailed anlysis of theseresults is given by Lammel and Nagel (2009) and Lammel etal. (2008).

Coping capacity in the case of a disastrous event encom-passes management and physical planning, social and eco-nomic capacity. This aims at pooling all different perspec-tives and results along the process chain in order to deriverecommendations for decision-makers before, during and af-ter an event. The basis for substantial management is realizedin a tool, where all results of the project are stored in a cen-tralized geodatabase. The tool streams all necessary data viainternet to an explicitly developed 3-D client browser with amultilingual interface. The browser enables the user, such asthe city administration of Padang, rescue teams, etc. to viewand analyze tsunami relevant data in 3-D at different levels ofdetail, even on the level of individual buildings. Because ofthe low data transfer rates in Indonesia’s IT infrastructure,it was essential to build on very efficient multi-resolutiontechniques for visualizing the DSM (Wahl et al., 2004) andfor efficiently mapping GIS data on it (Schneider and Klein,2007). These techniques further improve geometry repre-sentation (Wahl et al., 2008), texturing quality (Schneiderand Klein, 2008), and caching and streaming techniques insuch a way that the usability and efficiency of the viewer ap-plication within a restricted network infrastructure could beassured while retaining high rendering quality. The 3-D vi-sualization of geospatial data, adapted to the special require-ments of hazard management such as tsunami events, offersan effective tool even for inexperienced GIS users to interpretgeospatial data. This is indispensable if the results should bespread to a wider audience, such as stakeholder in Padang,with its limited computing and IT infrastructure. The mainte-nance and update of the geospatial data is very cost-efficientdue to its centralized data storage.

The tool intends to improve coping capacity in manifoldways: Since it combines the interdisciplinary research results


www.matsim.org


Fig. 14. Snapshots of the evacuation progress: Evacuation starts at 03:00 a.m. and the snapshots are taken after 1 min, 20 min and 40 min.The agents are colorized with respect to the time they need to evacuate. The evacuation time increases as the color moves from green toyellow to red.

level of individual buildings. Because of the low data transfer rates in Indonesia’s IT infrastructure, it was essential to build on very efficient multi-resolution techniques for visualizing the DSM (Wahl et al. 2004) and for efficiently mapping GIS data on it (Schneider & Klein 2007). These techniques further improve geometry representation (Wahl et al. 2008), texturing quality (Schneider & Klein 2008), and caching and streaming techniques in such a way that the usability and efficiency of the viewer application within a restricted network infrastructure could be assured while retaining high rendering quality. The 3D visualization of geospatial data, adapted to the special requirements of hazard management such as tsunami events, offers an effective tool even for inexperienced GIS users to interpret geospatial data. This is indispensable if the results should be spread to a wider audience, such as in Padang, with its limited computing and IT infrastructure. The maintenance and update of the geospatial data is very cost-efficient due to its centralized data storage.

Fig 15. Intuitive graphical user interface for data dissemination, developed by RSS in cooperation with the University of Bonn The tool intends to improve coping capacity in manifold ways: Since it combines the interdisciplinary research results and visualizes them in a consistent manner (building on the concepts described in Greß & Klein 2009), it enables to support decision making before, during and after a disastrous event. Before the event we aim at the correct assessment of various hazard scenarios to quantify exposed susceptible elements and people, assess their coping capacity and identify bottlenecks for evacuation. Furthermore the web application is used as platform for information dissemination and thus aims at awareness raising and guidance of political will. This analysis is the basis for recommendations in preparation for the expected disaster. During and after the tsunami event the platform is used to spatially plan and manage rescue measures. 5. Main findings and recommendations

Overall the interdisciplinary approach results in substantial spatial information covering the hazard impact, the exposed structures, the vulnerability of elements and groups of population, and the modeling of evacuation scenarios. The individual disciplines depended on and benefited from spatial information from other disciplines for the particular highly detailed analysis. Furthermore the manifold results enable a holistic perspective on processes in case of a hazard event. For example, through the interdisciplinary approach of the project it is possible to develop a microscopic evacuation simulation that not only relies on highly resolved geo-spatial data but also on a detailed socio-economic profile and highly resolved

Fig. 15. Intuitive graphical user interface for data dissemination, developed by RSS in cooperation with the University of Bonn.

and visualizes them in a consistent manner (building on theconcepts described in Greß and Klein, 2009), it enables tosupport decision making before, during and after a disastrousevent. Before the event we aim at the correct assessment ofvarious hazard scenarios to quantify exposed susceptible el-ements and people, assess their coping capacity and identifybottlenecks for evacuation. Furthermore the web applicationis used as platform for information dissemination and thusaims at awareness raising and guidance of political will. Thisanalysis is the basis for recommendations in preparation for

the expected disaster. During and after the tsunami event theplatform can be used to spatially plan and manage rescuemeasures.

5 Main findings and recommendations

Overall the interdisciplinary approach results in substantialspatial information covering the hazard impact, the exposedstructures, the vulnerability of elements and groups of pop-ulation, and the modeling of evacuation scenarios. The



individual disciplines depended on and benefited from spatialinformation from other disciplines for the particular highlydetailed analysis. Furthermore the manifold results enable aholistic perspective on processes in case of a hazard event.For example, through the interdisciplinary approach of theproject it is possible to develop a microscopic evacuationsimulation that not only relies on highly resolved geo-spatialdata but also on a detailed socio-economic profile and highlyresolved small scale inundation simulations. Without suchan interdisciplinary approach it would be hard to developsuch an all-inclusive simulation framework. The results letus quantify the impact of potential disastrous events, and thuslet us define and quantify counter-measures or relief actionsneeded.

The various scientific disciplines cannot perfectly map orsimulate the complexity of reality. But for decision makingbefore, during or after a disastrous event absolute accuracy isnot necessary. For example remote sensing cannot accuratelymap cadastral data regarding population, but it can provide afast assessment of the correct dimension of population andtheir whereabouts in the course of a day. Thus, the analysisis the crucial indication to identify weak points of a systemas basis to develop mitigation strategies and derive recom-mendations.

The main findings and recommendations in our studyaddress the questions we defined earlier in the conceptualoverview (cp. Sect. 2). 1) What kind of inundation dynamicsand what time frames are expected? 2) How many structuresand people and which groups of people are exposed for par-ticular hazard scenarios? 3) What factors influence differentresponses/behaviour to the tsunami early warning and copingcapacity of various social groups in emergency situation? 4)Which bottlenecks arise during the process of evacuation?5) Which scenario-specific decisions and recommendationshave to be met to minimize disaster risk?

1. The modeling of inundation scenarios using a multi-scenario approach reveal information about the affectedarea in the modeling domain for a given moment magni-tude. The level of risk that should be addressed by dis-aster managers should be determined before planningactivities could take place. Our approach capacitates lo-cal planers and stakeholders to decide where safe areascould be located e.g. for evacuation procedure. Infor-mation about maximum inundated areas are classifiedin four distinct probabilistic impact zones. Generally,zones of high risk for any earthquake triggered tsunamiare distributed along the coast but also along lower partsof river mouth and channels. Highly risky zones areseriously affected by water levels and flow velocitiesthat lead to complete damage of less robust houses.Houses and shelters in zones of medium risk will proba-bly withstand the impact, but flow path between housesand structures will be characterized by high flow veloc-ities that disable evacuation procedure completely. By

definition, zones of low risk will be damaged least bythe tsunami impact and the timeframe of inundation dy-namics will allow for the longest possible evacuationprogress. Since all results are still snapshots of work-in-progress, only qualitative statements could be givenat this stage, but we refer the reader to prospective pub-lications by the authors for quantitative conclusions anddetails.

2. The combination of the spatial impact of the potentialtsunami with the urban structures derived from remotelysensed data sets enables quantification of affected struc-tures, their types and their location. Exemplary 37 347structures were identified and localized in different riskzones of the hazard map. Furthermore, the land-coverclassification enables to identify and localize naturalsafe areas, defined as open spaces accessible by thestreet network larger than 10.000 m2 to be able to ac-commodate enough people in a rescue situation. Fur-thermore the area must be flat with land-cover bare soilor grassland and with respect to the hazard map a loca-tion outside any inundation area. The indirect deriva-tion of dynamic population distribution with respect tothe time of day results in the assessment of potentiallyaffected people and their location. Exemplary the over-lapping analysis of the spatial hazard scenario with the3-D city model and the population assessment resultedin 227 000 people at daytime and 217 000 at night-time.The socioeconomic analysis using the questionnaire re-veals the susceptibility and coping capacity of variousgroups and their locations.

3. Socio-economic vulnerability assessment throughquestionnaire-based surveying and additional qualita-tive studies on the specific themes such as access to thewarning and the potential response or behaviour of thepeople reveal the social bottlenecks to be consideredwhen establishing a people-centred effective earlywarning and evacuation planning. It provides betterunderstanding on how the people adapt to the currentrisk reduction measures and what kind of additionalmeasures are necessary. Spatial assessment of thesocio-economic parameters such as income, and otherrelated parameters enables projection of distributionof various social groups in the city areas as well as togenerate thematic maps on vulnerability. Furthermorethe usage of different media to warn people, e.g. TV,radio or sirens, clearly results in an increase of reachedpeople, and thus becomes a recommendation to bedeveloped further for warning dissemination. Theresults also clearly reveal that education programsshould focus on recommended evacuation routes.

4. Basically evacuation by foot is recommended due toheavy traffic jam situation in the last warning situationin 2007. From the pre-calculated evacuation simulation



it is possible to detect potential bottlenecks in advanceand give scenario specific evacuation recommendations.For example the bridge indicated in Fig. 14 is identi-fied as bottleneck with a too low capacity for the accu-mulating evacuees. In the case of the estimated mini-mum tsunami arrival time of 23 min a highly vulnera-ble spatial area is localized. Furthermore, one gets in-formation about estimated evacuation times on a geo-graphic highly resolved level. As shown in the exampleof Fig. 14 only after 40 min almost the entire populationwas evacuated from potentially inundated areas fromthe sample tsunami scenario. From this assessment, ar-eas where a vertical evacuation seems to be the onlypossible solution have to be localized. With the flexibil-ity of MATSim it will be possible to simulate differentscenarios regarding the distribution of the population,the inundation and the location of tsunami proof shel-ters. Therefore it will be possible to adapt the simulationif some parameters of the city change (e.g. widening ofnarrow streets, building of new bridges or the like).

5. The combination of the various aspects in the earlywarning chain enables to identify numerous recommen-dations and strategies for awareness raising, warningdissemination or spatial planning: The socioeconomicanalysis reveals that awareness of the people and house-hold characteristics have influence on the evacuation be-haviour. The correlation with remotely sensed data evenlets us localize extremely susceptible groups of peo-ple. Thus, we recommend standardized and location-specific education for the people, in order to make themaware on what to do in case of potential tsunami situ-ation and provision of an early warning; it is also im-portant to provide clear information to the people onwhere the recommended places and routes to evacuateare. Due to very short time availability to conduct evac-uation and limitation of various available media, it isalso recommended to empower the people to dissemi-nate the warning in an “organized” informal way, suchas through neighbourhood leaders or trained communityteams.

The evacuation simulation reveals bottlenecks – mainlyat bridges – giving manifold further options for improv-ing the coping capacity. We identified safe areas andrecommend to signpost them and with it to manage andthus release congestion areas in case of evacuation. Ifthe overall evacuation time for the situation as it is ex-ceeds an agreed maximum time (i.e., acceptance crite-rion), one possible strategy is to place shelters at loca-tions identified on the basis of simulation results. Wealso recommend improving accessibility of the two nat-ural elevation areas within the urban area of Padang asadditional vertical shelter areas. Next to bottlenecks, thewalking distance to a place of refuge is the second ma-jor determinant for the overall evacuation time. Addi-

tional shelters will reduce this walking distance for per-sons living or working close to those shelters. This willreduce the overall evacuation time, if for certain scenar-ios the individual evacuation time for many evacuees ismodelled as “too large” (i.e., higher than the acceptedtime). Therefore, this seems to be a very promisingstrategy. The success of this strategy can be quantifiedby performing a simulation with taking the shelters intoaccount.

6 Summary and outlook

The Last-Mile preparation for an expected natural disasteris an ongoing research project. The products developed aredisseminated to local stakeholders, decision-makers and of-ficials. Results serve as one essential information basis onhow to proceed with governance regarding hazard assess-ment, disaster mitigation, strategic planning and the formu-lation of policies on land use planning to reduce the potentialimpact of a tsunami at the city of Padang.

Future analysis will focus on tsunami impact assessmentutilizing vulnerability models of structures. Therefore theinterdisciplinary research combining civil engineering andremote sensing aims to correlate punctual survey data withthe area-wide physical parameters of the individual buildingfor extrapolation on the complete building stock of Padang.Thus, stable structures for vertical evacuation will be iden-tified. The integration of tsunami proof shelters will be animportant aspect regarding the evacuation simulation frame-work. Therewith, it will be possible to find appropriate loca-tions for those shelters and one could even assign shelters tothe people subjected to their whereabouts.

Interaction between social and engineering sciences pre-sented in this study offers a more comprehensive manner ofthe analysis and consequently a more applied product. Addi-tional disciplines or common efforts to establish more inten-sive participation of the local stakeholders may be necessary,especially in adjusting the format of the information systemto fit the local capacity and local planning purposes. Thiswould ensure the acceptance and sustainable application ofthe scientific product.

In addition it is planned to visualize simulation results inhigh resolution video clips, showing evacuation routes andother relevant information for education and awareness pur-poses. 3-D Studio Max will be used to render the videosequences. Wave simulation, evacuation routes and points,as well as instructions on behavior, are shown in the videosequences and will be used for awareness campaigns andtraining units. The video sequences are designed to allowrecognition of the city environment by using the 3-D citymodel textured with the building facade photos. In summary,the multifaceted interdisciplinary results within the “Last-Mile” project aim to holistically assess risk and vulnerabilityto mitigate human losses in case of the expected tsunamievent.



Acknowledgements.The authors would like to thank theDFG/BMBF special Programme “Geotechnologies” – EarlyWarning Systems in Earth Management. Sponsorship Code:03G0666A-E. We would also like to thank our partners fromAndalas University as well as the provincial and city authoritiesin the city of Padang, West Sumatra, Indonesia. We would alsolike to thank the Alfred-Wegener-Institute (AWI) for provision ofdeep-water propagation studies.

Edited by: S. TintiReviewed by: three anonymous referees

References

Barber, B.: Tsunami relief, US Agency for International Develop-ment, Washington DC, 28 pp., 2005.

Behrens, J.: TsunAWI – Unstructured Mesh Finite Element Modelfor the Computation of Tsunami Scenarios with Inundation, Pro-ceedings of NAFEMS Seminar, 2008.

Bernard, E. N., Mofjeld, H. O., Titov, V., Synolakis, C. E., and Gon-zalez, F. I.: Tsunami: Scientific frontiers, mitigation, forecastingand policy implications, Philos. T. R. Soc. A, 364, 1989–2007,2006.

Bernhard, E.: The U. S. National Tsunami Hazard Mitigation Pro-gram: A Successful State-Federal Partnership, Nat. Hazards, 35,5–24, 2005.

Bierlaire, M., Antonini, G., and Weber, M.: Behavioral dynamicsfor pedestrians, in: Moving through nets: The physical and socialdimensions of travel, edited by: Axhausen, K., Elsevier, 2003.

Birkmann, J.: Measuring vulnerability to promote disaster-resilientsocieties: Conceptual frameworks and definitions, in: MeasuringVulnerability to Natural hazards – Towards Disaster Resilient So-cieties, edited by: Birkmann, J., New York, United Nations Uni-versity, 9–54, 2006.

Birkmann, J., Dech, S., Goseberg, N., Lammel, G., Moder, F.,Nagel, K., Oczipka, M., Schlurmann, T., Setiadi, N., Siegert, F.,Strunz, G. and Taubenbock, H.: Numerical Last-Mile TsunamiEarly Warning and Evacuation Information System, in: Interna-tional Conference on Tsunami Warning (ICTW), Bali, Indonesia,12 pp., 12–14 November 2008a.

Birkmann, J., Setiadi, N., and Gebert, N.: Socio-economic vulner-ability assessment at the local level in context of tsunami earlywarning and Evacuation planning in the city of Padang, WestSumatra, Paper for International Conference on Tsunami Warn-ing, Nusa Dua, Indonesia, 12–14 November 2008b.

Blaikie, P., Cannon, T., Davis, I., and Wisner, B.: At Risk: Nat-ural Hazards, People’s Vulnerability and Disasters, Routeledge,London, S. 320, 1994.

Bogardi, J. and Birkmann, J.: Vulnerability Assessment: The FirstStep Towards Sustainable Risk Reduction, in: Disaster and So-ciety – From Hazard Assessment to Risk Reduction, edited by:Malzahn, D. and Plapp, T., Logos Verlag Berlin, Berlin, 75–82,2004.

Borner, A., Hirschmuller, H., Scheibe, K., Suppa, M., and Wohlfeil,J.: MFC – a modular line camera for 3D world modeling in Lec-ture Notes, in: Computer Science, Berlin/Heidelberg, Germany,January 2008.

Borrero, J. C., Sieh, K., Chlieh, M., and Synolakis, C. E.: Tsunamiinundation modeling for western Sumatra, P. Natl. Acad. Sci.

USA, 103, 19673–19677, available at:http://www.pnas.org,2006.

Borrero, J. C., Bourgeois, J., Harkins, G., and Synolakis, C. E.:How small-scale bathymetry affected coastal inundation in the1992 Nicaraguan tsunami, Proc. of the Fall AGU Meeting, SanFrancisco, 1997.

Burbridge, D., Cummins, P. R., Mleczko, R., and Thio, H. K.: AProbabilistic Tsunami Hazard Assessment for Western Australia,Pure Appl. Geophys., 165(11–12), 2059–2088(30), 2008.

Cannon, T., Twigg, J., and Rowell, J.: Social Vulnerability: Sus-tainable Livelihoods and Disasters, Report to DFID Conflict andHumanitarian Assistance Department (CHAD) and SustainableLivelihoods Support Office,http://www.proventionconsortium.org/themes/default/pdfs/CRA/DFIDSocialvulnerability.pdf,2003.

Cardona, O. D.: Environmental Management and Disaster Preven-tion: Two Related Topics: A Holistic Risk Assessment and Man-agement Approach, in: Natural Disaster Management, edited by:Ingleton, J., Tudor Rose, London, 1999.

Cardona, O. D.: Estimacion Holistica del Riesgo Sismico Uti-lizando Sistemas Dinamicos Complejos, Technical Universityof Catalonia, Barcelona,http://www.desenredando.org/public/varios/2001/ehrisusd/index.html, 2001

Chen, X. and Zhan, F.: Agent-based modeling and simulation ofurban evacuation: Relative effectiveness of simultaneous andstaged evacuation strategies. Paper 04-0329, Transportation Re-search Board Annual Meeting, Washington, DC, 2004.

Correll, R., Cramer, W., and Schellnhuber, H.-J.: Potsdam sustain-ability days, Symposium on “Methods and Models of Vulnera-bility Research, Analysis and Assessment”, 2001.

Davidson, R.: An Urban Earthquake Disaster Risk Index, The JohnA. Blume Earthquake Engineering Center, Department of CivilEngineering, Stanford University, Stanford, Report No. 121,1997.

Goseberg, N., Stahlmann, A., Schimmels, S., and Schlurmann, T.:Highly-resolved numerical modeling of tsunami run-up and in-undation scenario in the city of Padang, West Sumatra, Proc. ofthe 31st Int. Conference on Coastal Engineering, in press, 2009.

Goseberg, N. and Schlurmann, T.: Relevant factors on the extent ofinundation Based on Tsunami scenarios for the city of Padang,West Sumatra, Proc. of the International Conference on TsunamiWarning (ICTW), 2008.

Goto, C., Ogawa, Y., Shuto, N., and Imamura, N.: IOC Manual,UNESCO, No. 35, 1997.

Greß, A. and Klein, R.: Visualization Framework for the Integrationand Exploration of Heterogeneous Geospatial Data, Proceedingsof Eurographics 2009 – Areas Papers, Eurographics Association,27–34, April 2009.

Harig, S., Chaeroni, C., Behrens, J., and Schroeter, J.: TsunamiSimulations with unstructured grids (TsunAWI) and a com-parison to simulations with nested grids (Tsunami-N3), 6thInt. Workshop on Unstructured Mesh Numerical Modeling ofCoastal, Shelf and Ocean Flows, London, 2007.

Helbing, D., Farkas, I., and Molnar, P.: Simulation of pedestriancrowds in normal and evacuation situations, in: Pedestrian andEvacation Dynamics, Proceedings of the 1st international con-ference, Duisburg, 2001, Springer, 2002.

Hobeika, A. and Kim, C.: Comparison of traffic assignments inevacuation modeling, IEEE T. Eng. Manage., 45(2), 192–198,


http://www.pnas.org

http://www.proventionconsortium.org/themes/default/pdfs/CRA/DFIDSocialvulnerability.pdf

http://www.proventionconsortium.org/themes/default/pdfs/CRA/DFIDSocialvulnerability.pdf

http://www.desenredando.org/public/varios/2001/ehrisusd/index.html

http://www.desenredando.org/public/varios/2001/ehrisusd/index.html


1998.Intermap Technologies: Product Handbook & Quick Start Guide,

Standard Edition, v4.0, 2007.IPCC: Climate Change 2001: Impacts, Adaption, and Vulnerability,

Published for the Intergovernmental Panel on Climate Change,University Press, Cambridge, ISBN 0-521-80768-9, 2001.

Jha, M., Moore, K., and Pashaie, B.: Emergency evacuation plan-ning with microscopic traffic simulation, Transportation Re-search Board Annual Meeting, Washington, DC, Paper 04-2414,2004.

Kass, S., Post, J., Taubenbock, H., Roth, A., Strunz, G., Stotter, J.,Mardiatno, D., Ismail, F., and Anwar, H.: Automated assessmentof building vulnerability with high resolution IKONOS data forPadang, Indonesia, Proceedings of the International Symposiumon Disaster in Indonesia: Problems and Solutions, Padang, July2007.

Kl upfel, H., Meyer-Konig, T., Keßel, A., et al.: Simulat-ing evacuation processes and comparison to empirical results,in: Traffic and granular flow ’01, edited by: Fukui, M.,Sugiyama, Y., Schreckenberg, M., and Wolf, D., Springer,Berlin/Heidelberg/New York, 449–454, 2003.

Kwon, E. and Pitt, S.: Evaluation of emergency evacuation strate-gies for downtown event traffic using a dynamic network model,Transportation Research Board Annual Meeting, Washington,DC, Paper 05-2164, 2005.

Lammel, L., Rieser, M., and Nagel, K.: Large Scale MicroscopicEvacuation Simulation, Proc. 4th International Conference onPedestrian and Evacuation Dynamics, Wuppertal, Germany, inpress, 2009.

Lammel, L., Rieser, M., Nagel, K., Taubenbock, H., Strunz, G.,Goseberg, N., Schlurmann, T., Klupfel, H., Setiadi, N., and Birk-mann, J.: Emergency Preparedness in the case of a Tsunami –Evacuation Analysis and Traffic Optimization for the Indone-sian city of Padang, Proc. of the 4th Interantional Conferenceon Pedestrian and Evacuation Dynamics, Wuppertal, Germany,2008.

Lammel, L. and Nagel, K.: Multi agent based large-scale evacua-tion simulation, Transportation Research Board Annual Meeting,Washington, DC, Paper 09-2135, 2009.

Lammel, L., Klupfel, H., and Nagel, K.: Preliminary Result of aLarge Scale Microscopic Evacuation Simulation for the Cityof Padang in the Case of Tsunami, Proc. of the InternationalConference on Tsunami Warning (ICTW), Bali, Indonesia,http://pirba.ristek.go.id/ictw/e-proceedings/PDF/4.%20ICW/ICW05DE%20Laemmel%20Gregor.pdf, 2008.

Last-Mile Project:www.last-mile-evacuation.de, 2009.Lauterjung, J. and Rudloff, A.: GITEWS – The German Indone-

sian Tsunami Early Warning System, EOS T. Am. Geophys. Un.,86(52), Fall Meet. Suppl., 2005.

Leitmann, J.: Cities and Calamities: Learning from Post-Disasterresponse in Indonesia, in: Journal of Urban Health: Bulletin ofthe New York Academy of Medicine, 84(1), 144–153(10), 2007.

Leone, F., Denain, J.-C., Vinet, F., and Bachri, S.: Analysespatiale des dommages au bati d Banda Aceh (Sumatra, In-donesie): contributiona la connaissance du phenomene etal’ elaboration de scenarios de risque tsunami, in: Le tsunami du26 decembre 2004 en Indonesie – Rapport scientifique du pro-gramme TSUNARISQUE (2005–2006), 69–94, edited by: Lav-igne, F. and Paris, R., Delegation Interministerielle pour l’aide

Post-Tsunami (DIPT), Ambassade de France en Indonesie, Cen-tre National de la Recherche Scientifique, 356 pp., 2007.

Lynett, P., Wu, T.-R., and Liu, L.-F.: Modeling wave runup withdepth-integrated equations, Coast. Eng., 46, 89–107, 2002.

Marschalk, U., Roth, A., Eineder, M., and Suchandt, S.: Compari-son of DEMs derived from SRTM/X- and C-Band, in: Proceed-ings of the IGARSS 2004, Anchorage, USA, 2004.

McCarthy, J. J., Canziani, O. F., Leary, N. A., Dokken, D. J., andWhite, K. S. (Eds.): Climate Change: Impacts, Adaptation andVulnerability, Cambridge University Press, Cambridge, 2001.

McCloskey, J., Antonioli, A., Piatanesi, A., Sieh, K., Steacy, S.,Nalbant, S., Cocco, M., Giunchi, C., Huang, J. and Dunlop, P.:Tsunami threat in the Indian Ocean from a future megathrustearthquake west of Sumatra, Earth Planet. Sc. Lett., 265, 61–81,2008.

Moder, F., Oczipka, M., Siegert, F., Lehmann, F., Djajadihardaja, Y.S., Klein, R., and Wahl, R.: Last-Mile – Large-Scale topographicmapping of densely populated coasts in support of risk assess-ment of tsunami hazards, International Conference on TsunamiWarning, Bali, Indonesia, 12–14 November 2008.

Nielsen, O., Roberts, S., Gray, D., McPherson, A., and Hitchman,A.: Hydrodynamic modeling of coastal Inundation, in: MOD-SIM 2005 Int. Congress on Modeling and Simulation, Zerger, A.(Hrsg.), Modeling and Simulation Society of Australia and NewZealand, 518–523, 2005.

Papathoma, M., Dominey-Howes, D., Zong, Y., and Smith, D.: As-sessing tsunami vulnerability, an example from Herakleio, Crete,Nat. Hazards Earth Syst. Sci., 3, 377–389, 2003,http://www.nat-hazards-earth-syst-sci.net/3/377/2003/.

Pedersen, G., Liu, P. L., Yeh, H., and Synolakis, C.: Advanced nu-merical models for simulating tsunami waves and runup, WorldScientific, 10 pp., 2008.

Peeta, S. and Ziliaskopoulos, A.: Foundations of Dynamic TrafficAssignment: The Past, the Present and the Future, Netw. Spat.Econ., 1(3), 233–265, 2001.

Rastogi, B. K. and Jaiswal, R. K.: A Catalogue of Tsunamis inthe Indian Ocean, Science of Tsunami Hazards, 25(3), 128–143,2006.

Reese, S., Cousins, W. J., Power, W. L., Palmer, N. G., Tejakusuma,I. G., and Nugrahadi, S.: Tsunami vulnerability of buildings andpeople in South Java field observations after the July 2006 Javatsunami, Nat. Hazards Earth Syst. Sci., 7, 573–589, 2007,http://www.nat-hazards-earth-syst-sci.net/7/573/2007/.

Richter, R.: A spatially adaptive fast atmospheric correction algo-rithm, Int. J. Remote Sens., 17(6), S. 1201–1214, 1996.

Saita, J. and Nakamura, Y.: UrEDAS: The Early Warning Sys-tem for Mitigation of Disasters Caused by Earthquakes andTsunamis, in: Early Warning Systems for Natural Disaster Re-duction, edited by: Zschau, J. and Kuppers, N., Springer, 2003.

Schneider, M. and Klein, R.: Enhancing Textured Digital Eleva-tion Models Using Photographs. Proceedings of the Fourth Inter-national Symposium on 3D Data Processing, Visualization andTransmission (3DPVT’08), 2008.

Schneider, M. and Klein, R.: Efficient and Accurate Rendering ofVector Data on Virtual Landscapes, Journal of WSCG, 15(1–3),59–64, January 2007.

Schneiderbauer, S. and Ehrlich, D.: Risk, Hazard and People’s Vul-nerability to Natural Hazards: A Review of Definitions, Con-cepts and Data, European Commission-Joint Research Centre


http://pirba.ristek.go.id/ictw/e-proceedings/PDF/4.%20ICW/ICW05DE%20Laemmel%20Gregor.pdf

http://pirba.ristek.go.id/ictw/e-proceedings/PDF/4.%20ICW/ICW05DE%20Laemmel%20Gregor.pdf

www.last-mile-evacuation.de

http://www.nat-hazards-earth-syst-sci.net/3/377/2003/



(EC-JRC), Brussels, 2004.Schneiderbauer, S. and Ehrlich, D.: Social levels and hazard

(in)dependence in determining vulnerability, in: Measuring Vul-nerability to Natural hazards – Towards Disaster Resilient Soci-eties, edited by: Birkmann, J., New York, United Nations Uni-versity, 78–98, 2006.

Setiadi, N.: WP 1000: Socio-Economic Vulnerability Assessment– Numerical Last-Mile Tsunami Early Warning and EvacuationSystem, Presentation for the Status Meeting Geotechnologies,Osnabruck, 8 October 2008.

Sheffi, Y.: Urban Transportation Networks: Equilibrium Analysiswith Mathematical Programming Methods, Prentice-Hall, Engle-wood Cliffs, NJ, USA, 1985.

Sorensen, J. H.: Hazard Warning Systems: Review of 20 Years ofProgress, Natural Hazards Review, 1, 119–125, 2000.

Synolakis, C. E. and Bernard, E. N.: Tsunami science before andbeyond Boxing Day 2004, Philos. T. R. Soc. A, 364, 2231–2265,2006.

Taubenbock, H. and Roth, A.: A transferable and stable classifica-tion approach in various urban areas and various high resolutionsensors, Proc. of the Urban Remote Sensing Joint Event, Paris,France, S. 7, 2007.

Taubenbock, H., Roth, A., and Dech, S.: Linking structural urbancharacteristics derived from high resolution satellite data to pop-ulation distribution, in: Urban and Regional Data Management,Coors, Rumor, Fendel and Zlatanova (Hrsg.), Taylor & FrancisGroup, London, ISBN 978-0-41544059-2, S. 35–45, 2007.

Taubenbock, H., Post, J., Roth, A., Zosseder, K., Strunz, G., andDech, S.: A conceptual vulnerability and risk framework as out-line to identify capabilities of remote sensing, Nat. Hazards EarthSyst. Sci., 8, 409–420, 2008a,http://www.nat-hazards-earth-syst-sci.net/8/409/2008/.

Taubenbock, H., Post, J., Kiefl, R., Roth, A., Ismail, F., Strunz,G., and Dech, S.: Risk and vulnerability assessment to tsunamihazard using very high resolution satellite data, in: Proceedingsof the EARSeL Joint Workshop: Remote Sensing: New Chal-lenges of high resolution, edited by: Carsten, J., Bochum, Ger-many, ISBN 978-3-925143-79-3, 77–86,http://www.earsel.org/workshops/HighRes2008/Artikel/09Taubenboeck.pdf, 2008b.

Taubenbock, H., Post, J., Roth, A., Strunz, G., Kiefl, R., Dech, S.,and Ismail, F.: Multi-scale assessment of population distributionutilizing remotely sensed data – The case study Padang, WestSumatra, Indonesia, International Conference on Tsunami Warn-ing (ICTW), Bali, Indonesia, 12–14 November 2008c.

Taubenbock, H., Wurm, M., Setiadi, N., Gebert, N., Roth, A.,Strunz, G., Birkmann, J., and Dech, S.: Integrating Remote Sens-ing and Social Science – The correlation of urban morphologywith socioeconomic parameters, Urban Remote Sensing JointEvent, Shanghai, China, 2009.

Tinti, S., Manucci, A., Pagnoni, G., Armigliato, A., and Zaniboni,F.: The 30 December 2002 landslide-induced tsunamis in Strom-boli: sequence of the events reconstructed from the eyewitnessaccounts, Nat. Hazards Earth Syst. Sci., 5, 763–775, 2005,http://www.nat-hazards-earth-syst-sci.net/5/763/2005/.

Titov, V., Gonzalez, F., Bernhard, E., Eble, M., Mofjeld, H., New-man, J., and Venturato, A.: Real-time tsunami forecasting: Chal-lenges and solutions, Nat. Hazards, 35, 41–58, 2005.

Titov, V. V. and Synolakis, C. E.: Numerical modeling of tidal waverunup, J. Waterway Div.-ASCE, 124, 157–171, 1998.

Turner, B. L., Kasperson, R. E., Matson, P. A., McCarthy, J. J.,Corell, R. W., Christensen, L., Eckley, N., Kasperson, J. X.,Luers, A., Martello, M. L., Polsky, C., Pulsipher, A., and Schiller,A.: A framework for vulnerability analysis in sustainability sci-ence, P. Natl. Acad. Sci. USA, 100(14), 8074–8079, 2003.

United Nations International Strategy for Disaster Reduction(UN/ISDR): Living with Risk: A global Review of Disaster Re-duction Initiatives, United Nations International Strategy for Dis-aster Reduction, Geneva, Switzerland, UN Publications, 2004.

United Nations International Strategy for Disaster Reduc-tion (UN/ISDR): Global Survey of Early Warning Systems,UN/ISDR, ISDR/PPEW, 2006.

Wahl, R., Schnabel, R., and Klein, R.: From Detailed Digital Sur-face Models to City Models Using Constrained Simplification,Photogramm. Fernerkun., 3, 207–215, 2008

Wahl, R., Massing, M., Degener, P., Guthe, M., and Klein, R.: Scal-able Compression and Rendering of Textured Terrain Data, Jour-nal of WSCG, 12(3), 521–528, February 2004.

Zoppou, C. and Roberts, S.: Catastrophic Collapse of Water SupplyReservoirs in Urban Areas, J. Hydraul. Eng.-ASCE, 125, S. 686–695, 1999.



http://www.earsel.org/workshops/HighRes2008/Artikel/09_Taubenboeck.pdf

http://www.earsel.org/workshops/HighRes2008/Artikel/09_Taubenboeck.pdf


Documents

“Last-Mile” preparation for a potential disaster – Interdisciplinary